Trafficcompensated luminance measurement complementary …fudinfo.trafikverket.se/fudinfoexternwebb/Publikationer... · 2016-12-22 · the intra-day variation of the visibility of

REPORT

Contact person Date Reference Page

Mikael Lindgren 2016-04-08 MTk5P08986-2 1 (33) Measurement Technology +46 10 516 57 13 [email protected]

Trafikverket Att: Petter Hafdell Solna Strandväg 98 172 90 Sundbyberg

Traffic-compensated luminance measurement – complementary addition (2 appendices)

SP Technical Research Institute of Sweden

Postal address Office location Phone / Fax / E-mail This document may not be reproduced other than in full, except with the prior written approval of SP. SP

Box 857 SE-501 15 BORÅS Sweden

Västeråsen Brinellgatan 4 SE-504 62 BORÅS

+46 10 516 50 00 +46 33 13 55 02 [email protected]

Summary The document describes the findings from a study, performed as a complementary addition to previous work on Traffic compensated luminance estimation, as reported in MTk4P06467-2, Final report - Traffic compensated luminance estimation (2015-06-24).

During the test period the Lseq camera registered many scenes with extraordinary luminance distributions, caused by sun illuminated cloud edges and direct sun in the camera field of view. In the report, we discuss the influence of saturated pixels, field of view of the Lseq camera and the intra-day variation of the visibility of objects in the tunnel opening. Moreover, we show that the luminance value is highly dependent on the centring of the field of view. A rotation of the camera (or diode-based sensor) of approximately 1 degree will – for this particular scene and time of year – result in an average L20 luminance error of 5 percent.

One of the principal objectives of the original project was to estimate the impact of active traffic compensation to reduce the influence of traffic on road surface luminance estimation for lighting control. European standard EN 13201 dictates that the luminance should be measured from the road surface – and not the roof tops of the vehicles. In order to achieve this, we need to continuously keep track of the vehicles (using e.g. computer vision techniques) and make sure that, every time we sample a road patch, it is free from occluding vehicles.

The basic technology for active traffic compensation was evaluated in the original project. However, the technical evaluation was performed during a time of year when high-density traffic did not coincide with twilight or nocturnal conditions - the conditions of most interest for lighting control. In this complementary project, however, we have utilised the same set of camera prototypes to acquire data under highly relevant conditions.

The results show that, at peak traffic hours between 7am and 8am, the mean relative error between the compensated and uncompensated signals may exceed 60 %. The data clearly indicates a correlation between relative luminance errors and traffic flow - where high errors are more likely at times of high traffic flow. Based on these results, we believe that active traffic compensation should be a requirement to maintain errors at a reasonable level allowing lighting control while maintaining traffic safety.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 2 (33)


In addition to the above, we study the impact of various weather conditions on the road surface luminance. Under conditions with fog and/or snow, our results show that the average surface luminance may double thereby providing opportunities for significant energy savings through active dimming of light sources. Contributing authors: Kenneth Jonsson, Cipherstone Technologies AB Mikael Lindgren, SP Technical Research Institute of Sweden Björn Löfving, University of Gothenburg Rene Nilsson, Cipherstone Technologies AB Jörgen Thaung, University of Gothenburg

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 3 (33)


1. Background This document describes the results of the study financed by Trafikverket, as a complementary addition to the work reported in MTk4P06467-2 Final report – Traffic compensated luminance estimation1, dated 2015-06-24. The work presented here was proposed in the document ML150916-1rev1, entitled Projektförslag Trafikkompenserad luminansmätning komplettering2.

In the proposal, the following work was proposed:

Table 1. Description of work performed.

AP Work description Result Presentation

1.2 Additional data storage purchase

Complete The storage has been used for the additional image data collected.

1.6 Implement adaptive logging of image data

Complete An adaptive logging which collects image data more frequently when high Lseq values are noted has been implemented for the CIE 88 system

1.7 Collection and annotation of image data (EN 132015 and CIE 884)

Complete Image data has been collected and used for the evaluation reported here

4.7 Study the effect of changed sampling frequency when calculating Lseq

Complete The results are reported in section 2

4.9 Evaluate image material from fall 2015 in terms of Lseq


10.3 Evaluation of collected image data (EN 132015)


10.7 Search for particularly interesting lighting conditions. Analysis and visualization of these (EN 132015)


10.8 Calculation of relation between angular distance and pixel distance for an instrument placed above the road surface


11.1 Project management Complete

11.6 Presentation of results Complete Presented at Trafikverket 2016-01-13

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 4 (33)


2. Luminance data from Lseq - camera Data from the luminance camera looking into North entrance of Gnistängstunneln was collected during the period 2015-10-06 until 2015-11-21.

Lseq was calculated every 10 seconds and the intensity values for each segment in the Lseq grid were saved but the underlying camera images were only saved if Lseq values were high or the change in Lseq since last observation was big. Figure 1 shows Lseq observations when these criteria were fulfilled. Thus, for every Lseq observation in the graph, there is a corresponding luminance image available for further analysis.

By inspecting Figure 1 we find high daily maximum Lseq values on October 25 and November 8, 12, 14, 15, 18. Tone mapped images of the luminance distributions for these days are shown in Figure 2 and a summary of the conditions can be found in Table 2.

Table 2. Description of the scenarios with high Lseq observations. Lseq is calculated from luminance data with a saturation level of 225000 cd/m2.

Date Lseq Sky Sun Road

2015-10-25 427 10 % light clouds Direct Wet

2015-11-08 439 10 % light clouds Direct Dry


2015-11-14 461 10 % light clouds Direct Wet

2015-11-15 504 Clear sky Direct Dry


The luminance images were saturated at 225000 cd/m2, so all scenarios correspond to a higher Lseq value than given in Figure 1 and Table 2 because of saturated pixels. However, the measurement error is maybe not as big as one might first think, because the optical components of the camera will smear out the extremely high intensity levels from the sun to cover a larger area of the camera sensor chip. This imaging error will in some way compensate for the inability to measure the exact value. The luminance of the solar disc3 is probably in the range of 100 – 1000 Mcd/m2, but the diameter of the solar disc is about 0.5°, so the contribution to Lseq from a patch in one of the outer rings containing the sun, should not be unreasonably high. A rough estimation gives that the Lseq contribution from an outer patch containing the sun with an average solar disc luminance of 1000 Mcd/m2 could be about 800 cd/m2. During the test period the Lseq camera measured individual patch contributions in the range of 100 cd/m2, and total Lseq values of about 500 cd/m2 when direct sun was in the field of view. Therefore it is reasonable to assume that the measurement error estimates to a factor of just 2 – 3, even for extreme situations with direct sun in the field of view. The luminance images in Figure 2 are shown with different tone mapping for the upper and lower part. The upper part mainly contains the sky with high luminance levels, and the lower part contains lower luminance levels. By using different white points for the two areas, it is possible to see if there are clouds on the sky or where the sun is located in the upper part of the image and at the same time resolve contrasts in the lower part of the same image. Otherwise

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 5 (33)


the sky would have been white with no details visible or the lower part of the image would have appeared very dark.

Figure 1. Lseq observations during 2015-10-03 to 2015-11-21 at North entrance of Gnistängstunneln.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 6 (33)


Figure 2. Tone mapped luminance images of scenarios with high Lseq values.

2.1. Estimation of the visibility of an object in the tunnel opening A rough estimate of the visibility of an object in the tunnel opening is the ratio between the veiling luminance, Lseq, and the mean luminance in the 2° central ring, L2, of the Lseq grid. The higher luminance in the tunnel opening, the more visible will an object on the road become. However, the veiling luminance will reduce the contrasts of an object and the bigger the veiling luminance compared to the local mean luminance, the more it will influence the contrasts. Therefore, Lseq / L2 is an estimate of the visibility of an object.

Figure 3 shows the variation of Lseq / L2 over the test period. The maximum value is near 0.2 for almost all days. An example of the intraday variations is shown in Figure 4.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 7 (33)


By inspecting how the Lseq grid is located in the tunnel opening, c.f. Figure 8, it appears that L2 measures the luminance on the road about ten meters inside the tunnel. Therefore L2 is not exactly the same as the threshold luminance Lth.

Figure 3. The ratio Lseq / L2 estimates the visibility of an object in the tunnel opening and is here plotted for the whole test period.

Figure 4. An example of intraday variations of Lseq / L2 for October 5 and 6. By inspecting the exact time stamps it was found that the maximum values occurred 17:29 and 07:59, respectively.

The maximum value of ratio Lseq / L2 doesn’t coincide with the maximum value of Lseq, as shown in Figure 4 and in more detail in Figure 5 where Lseq / L2 graph is shown superimposed on the Lseq graph. Generally, it seems that Lseq / L2 is high in the morning and in the evening. In Figure 4, the maximum occurred at 17:29 on October 5 and at 07:59 on October 6. The luminance distributions at these times are shown in Figure 6. The same pattern repeats even on days with direct sun above the tunnel at noon, like 2015-11-08, c.f. Figure 2.

This observation indicates that the poorest visibility in the tunnel opening might occur for other luminance distributions than we might expect, i.e. not when the luminance levels are the highest at mid-day. One reason could be that for this particular tunnel portal, the roof is of raster type and that it works particularly well at mid-day luminance distributions. Further investigations are needed for full understanding of this phenomena, for example by investigating from where the road is illuminated at different times of the day.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 8 (33)


Figure 5. Variation of Lseq (solid line) on October 5 with the graph of ratio Lseq / L2 (dotted line) superimposed. The maximum value of Lseq / L2 does not coincide with maximum levels of Lseq.

Figure 6. Luminance images from two occasions on October 5 and 6 when the ratio Lseq / L2 was high.

2.2. Contribution from highest and lowest patches of Lseq grid In CIE 88:20044 page 11 there is a statement saying that the veiling luminance contributions from the two highest and the two lowest patches in the Lseq grid should not be taken into account because these patches are out of the field of vision. The reason is not given, but we might assume that the authors believe that the roof of the vehicle will block these parts of the field of view. However, modern cars tend to have large windscreens and trucks and buses have very little shielding from the roof in these areas, so it is far from certain that those patches should be excluded.

In an attempt to estimate the amount of contribution from these patches we calculated the fractional contribution to Lseq from the highest and lowers two patches respectively for all recorded luminance images during the test period. The result is presented in the histogram plots in Figure 7. We see that for most luminance distributions the two highest patches contribute with 10-15% to Lseq. The two lowest patches contribute with about 1-3% to Lseq and are more or less negligible. For some luminance distributions the two highest patches

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 9 (33)


contribute to over 40% of Lseq, which turned out to be because of direct sun, or illuminated cloud edges, in these areas, see Figure 8. Two other examples of luminance distributions where the contribution to Lseq from the highest patches are 25% and 10% respectively, are also given in Figure 8.

Figure 7. Histogram showing the contribution from the two highest patches (left) and the two lowest patches (right) in the Lseq grid.

Figure 8. Some examples of luminance distribution scenarios when the contributions to Lseq from the two highest patches in the Lseq grid vary between 47% - 10%.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 10 (33)


3. Centring the field of view As dictated in CIE 88:20044, the L20 field of view should be centred vertically at 25 percent of the distance from the road surface to the tunnel ceiling. Similarly, the Lseq field of view should be centred vertically at 50 percent of the distance. In the original TKLM study, we investigated the impact of the centring of the L20 field-of-view on the luminance values. We concluded that shifting the L20 field of view from 25 to 50 percent of the distance (i.e. using the Lseq centre point as opposed to the L20 centre point) would offset the luminance values by 5 percent on average over a longer period of time. In this study, the re-centring of the L20 measurement region was performed by shifting the L20 circle within the image region. Below we compute the angle corresponding to the shift distance in pixels for Test Site 2.

The relationship between the pixel shift and the camera rotation angle is expressed as follows:

β=arctan �2𝑑𝑦𝑡𝑡𝑡 �

𝛼2��

where α is the camera aperture, d is the pixel shift in the image space, and y is the image height in pixels. These distances and angles are illustrated in Figure 9.

Figure 9. Relationship between pixel shift and camera rotation angle.

For Test Site 1, we have d = 12 pixels, y = 600 pixels and α = 47.4 degrees. Thus, we get β = 1.01 degrees. In other words, a rotation of the camera (or diode-based sensor) of approximately 1 degree will – for this particular scene and time of year – result in an average L20 luminance error of 5 percent.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 11 (33)


4. Road Surface Luminance Data Analysis In this section, we present an analysis of the road surface luminance data acquired at TKLM Test Site 2 where EN 13201 measurements have been performed during October and November 2015, and January 2016. This includes some overall statistics, a categorisation of the measurement periods, a study of the impact of traffic compensation on luminance estimation, and an investigation into the influence of various weather conditions on the road surface luminance.

Some of the key observations made in this analysis are:

• At peak traffic hours between 7am to 8am, the highest mean relative errors between the compensated and uncompensated luminance signals exceed 50 % in October and 60 % in November.

• The data clearly indicates a correlation between relative luminance errors and traffic flow – where high errors are more likely at times of high traffic flow.

• In our study, as expected, we cannot see a clear relationship between relative luminance error and luminance level, e.g. a high error does not necessarily correspond to a high (or low) luminance level.

• In general, the road surface luminance is higher in the left lane than in the right lane. This is probably due to the positioning of the light sources in the centre of the road.

• In general, the relative luminance error is higher in the right lane compared to the left lane. The reason is probably a combination of higher average traffic flow and lower road surface luminance in the right lane.

• In wet conditions, the luminance level is significantly higher in the left lane than the right lane. This may be due to the positioning of the light sources, the viewing angle of the camera and the relative abilities of the two lanes to drain water from the road surface.

• Fog can cause an increase in the average road surface luminance of 50 percent. Fog in combination with snowy conditions can double the road surface luminance.

• During the measurement period, the lowest mean luminance was 0.4 cd/m2 – far below the 1 cd/m2 requirement for the road class ME3. The typical mean luminance during a dry night was around 0.5 – 0.6 cd/m2. However, we are probably underestimating the road surface luminance value due to the greater angle between the camera optical axis and the road surface compared to the standard observer.

4.1. General measurement statistics In Table 3 and Table 4, we list road surface luminance error statistics for October 2015. We show the mean and median relative errors between the traffic compensated and uncompensated luminance values. The relative error ε between a traffic compensated value Lt and the corresponding uncompensated value Lu is computed as follows:

𝜀 = (𝐿t − 𝐿u)/𝐿t

Also, we show the standard deviation of the luminance errors, and the minimum/maximum errors. In Table 3, we show error statistics for the time period 7pm to 7am and for the right lane only. In Table 4, we show the corresponding statistics for the morning traffic between 7am and 8am. In both tables, the three highest mean errors are highlighted in red. For some nights (e.g. 2015-10-19 – 2015-10-20) , measurements are not available due to technical problems.

As can be seen in Table 3, the highest mean error between 7pm and 7am exceeds 15 %. In the high-density period between 7am and 8am, the highest mean error exceeds 50 %.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 12 (33)


Table 3. October relative luminance error ϵ, 7pm – 7am, right lane.

Measurement Night Mean Median Std dev Min Max 2015-10-01 – 2015-10-02 0,1220 0,0169 0,2766 0,0000 2,6660 2015-10-02 – 2015-10-03 0,0823 0,0161 0,1834 0,0000 1,6191 2015-10-03 – 2015-10-04 0,0973 0,0164 0,2294 0,0000 3,4208 2015-10-04 – 2015-10-05 0,1154 0,0109 0,4274 0,0000 9,5896 2015-10-05 – 2015-10-06 0,1170 0,0147 0,2472 0,0000 2,7510 2015-10-06 – 2015-10-07 0,1193 0,0147 0,2550 0,0000 2,5299 2015-10-07 – 2015-10-08 0,1342 0,0164 0,2912 0,0000 2,9802 2015-10-08 – 2015-10-09 0,1407 0,0314 0,3416 0,0000 3,1946 2015-10-09 – 2015-10-10 0,0906 0,0152 0,1820 0,0000 2,1078 2015-10-10 – 2015-10-11 0,0950 0,0164 0,1951 0,0000 1,5773 2015-10-11 – 2015-10-12 0,1282 0,0111 0,2804 0,0000 2,2011 2015-10-12 – 2015-10-13 0,1498 0,0140 0,4484 0,0000 8,4536 2015-10-13 – 2015-10-14 0,1474 0,0169 0,3415 0,0000 3,7199 2015-10-14 – 2015-10-15 0,1450 0,0188 0,2931 0,0000 2,0898 2015-10-15 – 2015-10-16 0,1414 0,0186 0,2803 0,0000 2,0941 2015-10-16 – 2015-10-17 0,1069 0,0150 0,2839 0,0000 3,2283 2015-10-17 – 2015-10-18 0,0874 0,0185 0,1724 0,0000 1,4384 2015-10-18 – 2015-10-19 0,1394 0,0124 0,3123 0,0000 3,2010 2015-10-19 – 2015-10-20 N/A N/A N/A N/A N/A 2015-10-20 – 2015-10-21 N/A N/A N/A N/A N/A 2015-10-21 – 2015-10-22 N/A N/A N/A N/A N/A 2015-10-22 – 2015-10-23 0,1541 0,0334 0,3627 0,0000 3,2098 2015-10-23 – 2015-10-24 0,1073 0,0171 0,2281 0,0000 1,6557 2015-10-24 – 2015-10-25 0,0659 0,0213 0,1693 0,0000 1,9866 2015-10-25 – 2015-10-26 0,1254 0,0114 0,2764 0,0000 2,2081 2015-10-26 – 2015-10-27 0,1157 0,0129 0,2798 0,0000 2,6588 2015-10-27 – 2015-10-28 0,1137 0,0172 0,2309 0,0000 1,7656 2015-10-28 – 2015-10-29 N/A N/A N/A N/A N/A 2015-10-29 – 2015-10-30 0,0851 0,0174 0,1543 0,0001 0,9915

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 13 (33)


Table 4. October relative luminance error ϵ, 7am - 8am, right lane.

Measurement Night Mean Median Std dev Min Max 2015-10-01 – 2015-10-02 0,1569 0,1070 0,1510 0,0000 0,8253 2015-10-02 – 2015-10-03 0,0479 0,0395 0,0542 0,0000 0,4263 2015-10-03 – 2015-10-04 0,0429 0,0355 0,0412 0,0000 0,2670 2015-10-04 – 2015-10-05 0,1022 0,0776 0,0870 0,0000 0,4179 2015-10-05 – 2015-10-06 0,1358 0,0949 0,1241 0,0000 0,5563 2015-10-06 – 2015-10-07 0,1538 0,1061 0,1495 0,0000 0,7232 2015-10-07 – 2015-10-08 0,1942 0,1156 0,2823 0,0000 1,3626 2015-10-08 – 2015-10-09 0,2588 0,2075 0,2821 0,0000 1,7243 2015-10-09 – 2015-10-10 0,0473 0,0363 0,0478 0,0000 0,3697 2015-10-10 – 2015-10-11 0,0435 0,0257 0,0920 0,0000 0,6922 2015-10-11 – 2015-10-12 0,2021 0,1423 0,2087 0,0000 1,2246 2015-10-12 – 2015-10-13 0,2117 0,1143 0,2361 0,0000 1,0828 2015-10-13 – 2015-10-14 0,3075 0,1776 0,4309 0,0000 2,2595 2015-10-14 – 2015-10-15 0,2542 0,1613 0,3081 0,0000 1,6865 2015-10-15 – 2015-10-16 0,3502 0,1841 0,4652 0,0000 2,0205 2015-10-16 – 2015-10-17 0,1178 0,0411 0,2638 0,0000 1,4229 2015-10-17 – 2015-10-18 0,1197 0,0439 0,3895 0,0000 2,1774 2015-10-18 – 2015-10-19 0,4563 0,2545 0,6041 0,0000 2,5803 2015-10-19 – 2015-10-20 N/A N/A N/A N/A N/A 2015-10-20 – 2015-10-21 N/A N/A N/A N/A N/A 2015-10-21 – 2015-10-22 N/A N/A N/A N/A N/A 2015-10-22 – 2015-10-23 0,5317 0,3317 0,5904 0,0000 3,2465 2015-10-23 – 2015-10-24 0,1395 0,0539 0,3426 0,0000 1,9255 2015-10-24 – 2015-10-25 0,0358 0,0318 0,0246 0,0000 0,1081 2015-10-25 – 2015-10-26 0,0981 0,0930 0,0648 0,0000 0,2926 2015-10-26 – 2015-10-27 0,1335 0,1078 0,0977 0,0000 0,4455 2015-10-27 – 2015-10-28 0,1536 0,1000 0,1597 0,0000 0,8324 2015-10-28 – 2015-10-29 N/A N/A N/A N/A N/A 2015-10-29 – 2015-10-30 N/A 0,0850 N/A 0,0000 N/A

In Table 5 and Table 6 we list road surface luminance error statistics for November 2015. As for October, we show the mean and median relative errors between the traffic compensated and uncompensated luminance values. Also, we show the standard deviation of the luminance errors, and the minimum/maximum errors. In Table 5, we show error statistics for the time period 5pm to 7am and for the right lane only. In Table 6 we show the corresponding statistics for the morning traffic between 7am and 8am. In both tables, the three highest mean errors are highlighted in red. For one night (2015-11-11 – 2015-11-12) , measurements are not available due to technical problems.

As can be seen in Table 5, the highest mean error between 5pm and 7am exceeds 10 %. In the high-density period between 7am and 8am, the highest mean error exceeds 60 %.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 14 (33)


Table 5. November relative luminance error ϵ, 5pm – 7am, right lane.

Measurement Night Mean Median Std dev Min Max 2015-11-01 – 2015-11-02 0,0599 0,0155 0,1097 0,0000 0,7543 2015-11-02 – 2015-11-03 0,0969 0,0198 0,1650 0,0000 1,0504 2015-11-03 – 2015-11-04 0,0990 0,0262 0,1784 0,0000 2,0055 2015-11-04 – 2015-11-05 0,0913 0,0219 0,1448 0,0001 0,8290 2015-11-05 – 2015-11-06 0,0939 0,0392 0,1551 0,0001 2,0420 2015-11-06 – 2015-11-07 0,0906 0,0357 0,1553 0,0001 2,5990 2015-11-07 – 2015-11-08 0,0855 0,0370 0,1374 0,0000 1,3401 2015-11-08 – 2015-11-09 0,0967 0,0316 0,1555 0,0000 0,9001 2015-11-09 – 2015-11-10 0,0781 0,0214 0,1391 0,0000 1,0125 2015-11-10 – 2015-11-11 0,0980 0,0326 0,1765 0,0000 2,1006 2015-11-11 – 2015-11-12 N/A N/A N/A N/A N/A 2015-11-12 – 2015-11-13 0,0895 0,0264 0,1514 0,0000 1,3563 2015-11-13 – 2015-11-14 0,0880 0,0328 0,1516 0,0001 1,4310 2015-11-14 – 2015-11-15 0,0740 0,0180 0,1379 0,0000 0,8910 2015-11-15 – 2015-11-16 0,0771 0,0159 0,1430 0,0000 1,2006 2015-11-16 – 2015-11-17 0,0721 0,0258 0,1265 0,0000 1,2631 2015-11-17 – 2015-11-18 0,0982 0,0290 0,1689 0,0000 2,2075 2015-11-18 – 2015-11-19 0,0944 0,0401 0,1404 0,0000 1,2043 2015-11-19 – 2015-11-20 0,1087 0,0287 0,1728 0,0000 1,1254 2015-11-20 – 2015-11-21 0,0878 0,0186 0,1602 0,0000 1,1053 2015-11-21 – 2015-11-22 0,0743 0,0178 0,1432 0,0000 1,1836 2015-11-22 – 2015-11-23 0,0770 0,0172 0,1395 0,0000 1,0985 2015-11-23 – 2015-11-24 0,0663 0,0206 0,1377 0,0000 1,5300 2015-11-24 – 2015-11-25 0,1094 0,0479 0,1675 0,0000 1,5090 2015-11-25 – 2015-11-26 0,0910 0,0315 0,1432 0,0000 1,3411 2015-11-26 – 2015-11-27 0,0723 0,0266 0,1123 0,0001 1,1229 2015-11-27 – 2015-11-28 0,0813 0,0377 0,1251 0,0000 0,9817 2015-11-28 – 2015-11-29 0,0768 0,0349 0,1212 0,0000 0,9778 2015-11-29 – 2015-11-30 0,0859 0,0414 0,1260 0,0001 1,1625

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 15 (33)


Table 6. November relative luminance error ϵ, 7am - 8am, right lane.


REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 16 (33)


The luminance recorded for the period is reported in Table 7 – Table 8. The lowest values are highlighted in red.

Table 7. October luminance statistics (traffic compensated, Lt), 7pm - 7am, right lane.

Measurement Night Mean Median Std dev Min Max 2015-10-01 – 2015-10-02 0,5804 0,5493 0,2557 0,2954 3,8807 2015-10-02 – 2015-10-03 0,5620 0,5439 0,2357 0,2440 3,1892 2015-10-03 – 2015-10-04 0,5426 0,5563 0,1361 0,2480 1,6222 2015-10-04 – 2015-10-05 0,5554 0,5566 0,1216 0,2342 1,7116 2015-10-05 – 2015-10-06 0,5489 0,5444 0,0955 0,2453 0,9434 2015-10-06 – 2015-10-07 0,5478 0,5474 0,0937 0,2837 0,8295 2015-10-07 – 2015-10-08 0,5501 0,5510 0,0943 0,2653 0,9150 2015-10-08 – 2015-10-09 0,4021 0,4072 0,0909 0,1665 0,5925 2015-10-09 – 2015-10-10 0,5155 0,5499 0,1165 0,2412 0,6781 2015-10-10 – 2015-10-11 0,5141 0,5408 0,1193 0,2409 0,6608 2015-10-11 – 2015-10-12 0,5296 0,5403 0,1015 0,2621 1,1704 2015-10-12 – 2015-10-13 0,5952 0,5487 0,7277 0,2873 12,8410 2015-10-13 – 2015-10-14 0,5341 0,5346 0,1361 0,2929 1,8320 2015-10-14 – 2015-10-15 0,5356 0,5423 0,0982 0,2325 0,7653 2015-10-15 – 2015-10-16 0,5334 0,5423 0,1010 0,2704 1,1978 2015-10-16 – 2015-10-17 0,5216 0,5480 0,1143 0,2489 0,6605 2015-10-17 – 2015-10-18 0,5037 0,5329 0,1204 0,2254 0,6556 2015-10-18 – 2015-10-19 0,5079 0,5202 0,0993 0,2224 0,8217 2015-10-19 – 2015-10-20 N/A N/A N/A N/A N/A 2015-10-20 – 2015-10-21 N/A N/A N/A N/A N/A 2015-10-21 – 2015-10-22 N/A N/A N/A N/A N/A 2015-10-22 – 2015-10-23 0,4204 0,4470 0,0934 0,1656 0,5904 2015-10-23 – 2015-10-24 0,5111 0,5391 0,1148 0,2346 0,7427 2015-10-24 – 2015-10-25 0,4604 0,4543 0,1284 0,1078 1,5236 2015-10-25 – 2015-10-26 0,5555 0,5470 0,1567 0,2811 2,1755 2015-10-26 – 2015-10-27 0,5265 0,5384 0,1128 0,2749 1,5432 2015-10-27 – 2015-10-28 0,4827 0,4858 0,1063 0,1452 0,9158 2015-10-28 – 2015-10-29 N/A N/A N/A N/A N/A 2015-10-29 – 2015-10-30 0,6176 0,5887 0,0981 0,5436 1,4870

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 17 (33)


Table 8. November luminance statistics (traffic compensated, Lt), 5pm - 7am, right lane.


4.2. Categorisation of the measurement periods In Table 9, we categorise each of the nights in October and November based on the behaviour of the luminance data. The categories are: “Nothing To Report” (NTR), “Rain”, “Dark” and “Spike”. The first two categories are self-explanatory. The category “Dark” indicates a night with lower than normal luminance levels. The category “Spike” indicates significant deviations in the luminance signal at specific points in time. The reader is referred to the full listing of luminance graphs in the appendices for more detailed information. In the following sections, we show some highlights from these measurement periods illustrating the different categories. The entries shown in red colour in Table 9 are illustrated in the following sections.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 18 (33)


Table 9. Categorization of measurement periods, October and November.

Measurement Night Days October November 2015-1X-01 – 2015-1X-02 Thu-Fri NTR Rain 2015-1X-02 – 2015-1X-03 Fri-Sat NTR NTR 2015-1X-03 – 2015-1X-04 Sat-Sun NTR Spike @ 8.20pm 2015-1X-04 – 2015-1X-05 Sun-Mon NTR NTR 2015-1X-05 – 2015-1X-06 Mon-Tue NTR Rain 2015-1X-06 – 2015-1X-07 Tue-Wed NTR Rain 2015-1X-07 – 2015-1X-08 Wed-Thu NTR Rain 2015-1X-08 – 2015-1X-09 Thu-Fri Rain Rain in morning 2015-1X-09 – 2015-1X-10 Fri-Sat Spike @ 9.40pm Rain 2015-1X-10 – 2015-1X-11 Sat-Sun NTR Rain 2015-1X-11 – 2015-1X-12 Sun-Mon NTR NTR 2015-1X-12 – 2015-1X-13 Mon-Tue Spike @ 8.20pm NTR 2015-1X-13 – 2015-1X-14 Tue-Wed Spike @ 8.30pm Rain 2015-1X-14 – 2015-1X-15 Wed-Thu NTR NTR 2015-1X-15 – 2015-1X-16 Thu-Fri NTR NTR 2015-1X-16 – 2015-1X-17 Fri-Sat NTR Rain 2015-1X-17 – 2015-1X-18 Sat-Sun NTR Rain in morning 2015-1X-18 – 2015-1X-19 Sun-Mon NTR Rain 2015-1X-19 – 2015-1X-20 Mon-Tue NTR NTR 2015-1X-20 – 2015-1X-21 Tue-Wed NTR NTR 2015-1X-21 – 2015-1X-22 Wed-Thu NTR NTR 2015-1X-22 – 2015-1X-23 Thu-Fri Rain NTR 2015-1X-23 – 2015-1X-24 Fri-Sat Spike @ 6am Rain 2015-1X-24 – 2015-1X-25 Sat-Sun Rain Rain 2015-1X-25 – 2015-1X-26 Sun-Mon NTR Rain in morning 2015-1X-26 – 2015-1X-27 Mon-Tue NTR NTR 2015-1X-27 – 2015-1X-28 Tue-Wed NTR Rain 2015-1X-28 – 2015-1X-29 Wed-Thu NTR Rain 2015-1X-29 – 2015-1X-30 Thu-Fri NTR Rain

4.3. The impact of traffic compensation To investigate the impact of traffic flow on luminance estimation in detail, we plotted the relative luminance error as a function of traffic flow for one morning in November. The traffic flow data was obtained from the regional database of the National Transport Administration and expresses the number of vehicles per hour. The resulting plot is shown in Figure 10. As can be seen in the figure, there is a correlation between luminance errors and traffic flow – high errors are more likely to appear in high flow. On this particular morning from 5am to 8am, all errors above 50 % occur at traffic flows above 1000.

However, it is difficult to study the relationship between luminance errors and traffic flow in isolation since, at this time of day, the luminance level is also changing rapidly. Since the traffic compensation algorithm has a built-in inertia meaning that it will not adapt directly to

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 19 (33)


quickly changing light levels, the error between the compensated and uncompensated signals will be partly due to traffic and partly due to the slower adaptation rate of the compensated signal.

Figure 10. Relative luminance error ϵ as function of traffic flow, 2015-11-24 5am to 8am.

We have also investigated the relationship between relative luminance error and traffic compensated luminance level. This is shown in Figure 11. From this measurement period, it is difficult to draw conclusions regarding the relationship – a high error does not necessarily correspond to a high luminance level.

Figure 11. Relative luminance error ϵ as function of road surface luminance Lt (cd/m2), 2015-11-24 5am to 8am.

0 200 400 600 800 1000 1200 1400 1600 1800 20000,0000

0,5000

1,0000

1,5000

2,0000

2,5000

Traffic Flow (number of vehicles per hour)

Rela

tive

Lum

inan

ce E

rror

ϵ

0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 1,4000 1,6000 1,80000,0000

0,5000

1,0000

1,5000

2,0000

2,5000

Traffic Compensated Road Surface Luminance Lt

Rela

tive

Lum

inan

ce E

rror

ϵ

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 20 (33)


In Figure 12, we show the mean road surface luminance (compensated and uncompensated) and the relative error as functions of time for this measurement period. In Figure 13, we show the traffic flow for the same period. As can be seen in Figure 13, the traffic flow is steadily increasing. The light levels starts rising from just before 7am. The intense traffic in combination with quickly rising light levels result in large deviations between the compensated and uncompensated signals.

Figure 12. Mean luminance (traffic compensated Lt in dark blue, uncompensated Lu in green, cd/m2) and relative error ϵ (red, no unit) as functions of time of day, 2015-11-24 5am to 8am.

Figure 13. Traffic flow as function of time of day, 2015-11-24 5am to 8am.

05:00:00AM

06:00:00AM

07:00:00AM

08:00:00AM

0,0000

0,5000

1,0000

1,5000

2,0000

2,5000

3,0000

3,5000

4,0000

4,5000

Time of Day

Mea

n Lu

min

ance

(cd/

m2)

and

Rel

ativ

e Er

ror ϵ

05:00:00AM

06:00:00AM

07:00:00AM

08:00:00AM

0

200

400

600

800

1000

1200

1400

1600

1800

2000

Time of Day

Traf

fic F

low

(num

ber o

f veh

icle

s per

hou

r)

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 21 (33)


In an attempt to isolate the relationship between relative luminance error and traffic flow, we selected another day and time period where the light levels are more or less constant. This is shown in Figure 14. During the chosen three hour period the light level was between 1.0 and 1.1 cd/m2 all the time. Also in this figure the high errors appear more common at high traffic flows.

Figure 14. Relative luminance error ϵ as function of traffic flow, 2015-11-23 4.30pm to 7.30pm

In Figure 15, we show the relationship between relative luminance error and traffic flow for a third day and time period. In this case, it is a Saturday evening between 5pm and 8pm and the traffic flow is significantly lower than in previous time periods. At this low level of traffic flow, the relationship between luminance error and flow is less clear.

0 200 400 600 800 1000 1200 1400 1600 18000

0,2

0,4

0,6

0,8

1

1,2

1,4


Rela

tive

Lum

inan

ce E

rror

ϵ

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 22 (33)


Figure 15. Relative luminance error ϵ as function of traffic flow, 2015-11-14 5pm to 8pm.

4.4. Relative luminance error distributions In Figure 16, we show the distributions of relative luminance errors for one night (October 4 to 5) and for the left and right lanes, separately. The summary statistics for the two periods 7pm to 7am and 7am to 8am are also shown in the figure. As can be seen in the figure, most of the errors are relatively small but the distribution has a long tail. These error distributions are representative for many of the nights during October and November.

Note that the right lane errors are higher than the left lane errors which is probably due to higher average traffic flow in the right lane.

0 200 400 600 800 1000 12000

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9


Rela

tive

Lum

inan

ce E

rror

ϵ

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 23 (33)


Figure 16. Relative luminance error distributions, October 4 to 5. Top row is left lane and bottom row is right lane.

4.5. The impact of weather conditions In Figure 17, we show an example of a night where it starts to rain during the night. As can be seen in the figure, the luminance level is significantly higher in the left lane than the right lane when the road surface is wet. This can also be seen in the luminance images shown in Figure 18 – the left one was taken before the rain and the right one during the rain. This may be due to the positioning of the light sources, the viewing angle of the camera and the relative abilities of the two lanes to drain water from the road surface.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 24 (33)


Note that the EN 13201-system uses high-dynamic range images composed from two exposure times but, in Figure 18, we only show a single exposure time (where some of the image regions are saturated).

Figure 17. EN 13201 average surface luminance: November 25, 5pm, to November 26, 8am. Top row is left lane and bottom row is right lane. Uncompensated Lu in red, compensated Lt in blue.

Left lane

Right lane

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 25 (33)


Figure 18. Images taken 2015-11-25 at 12pm (left) and 2015-11-26 at 3am (right). One out of two exposure times only.

In Figure 19, we show another example of a rainy night. Luminance images just before and after the sharp rise in mean road surface luminance at just after 9pm on November 16 are shown in Figure 20.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 26 (33)


Figure 19. EN 13201 average surface luminance: November 16, 5pm, to November 17, 8am. Top row is left lane and bottom row is right lane. Uncompensated Lu in red, compensated Lt in blue.

Figure 20. Images taken at 9.00pm (left) and 9.20pm (right) on 2015-11-16. One out of two exposure times only.

In Figure 21 and Figure 22, we show an example of a mixed night involving several rain showers. The night starts dry, the street lights are turned on at around 8.30pm and a first rain

Left lane

Right lane

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 27 (33)


shower appears at around 10.30pm. Then, the road surface slowly dries up and rain appears again at around 2am. In Figure 22, we show images just before and after some of the dramatic changes in the luminance values.

Figure 21. EN 13201 average surface luminance: October 24, 6pm, to October 25, 8am. Top row is left lane and bottom row is right lane. Uncompensated Lu in red, compensated Lt in blue.

Right lane

Left lane

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 28 (33)


Figure 22. Images taken at 8.43pm (top left) and 10.43pm (top right) on 2015-10-24, and 01.43am (bottom left) and 02.23am (bottom right) on 2015-10-25. One out of two exposure times only.

In Figure 23 and Figure 24, we show a night involving several events of interest. The first event occurs between 8pm and 10pm on 2016-01-20 as can be seen in Figure 23. The sharp rise in luminance values is due to road works and is not related to any change in weather condition. The second event, however, is due to fog slowly building up in the field of view. In Figure 24, we show four images taken from the start of the fog until it has cleared from the field of view of the camera. The average road surface luminance increases from around 0.7 cd/m2 to 1.1 cd/m2 in the left lane – an increase of around 50 percent.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 29 (33)


Figure 23. EN 13201 average surface luminance: January 20, 5pm, to January 21, 8am. Left lane (blue) and right lane (red), with traffic compensation.

Figure 24. Images taken at 4.00am (top left), 4.20am (top right), 4.40am (bottom left) and 5.00am (bottom right) on 2016-01-21. One out of two exposure times only.

Another night involving both fog and snowy conditions is the night between 2016-01-23 and 2016-01-24 where the luminance values are consistently high. The average road surface luminance as a function of time is shown in Figure 25 and a selection of images distributed through the night is shown in Figure 26. The fog is particularly dense at around midnight. In the left lane, the average luminance is around 1.5 cd/m2 all night which is around twice the average value a typical dry-surface night.

0

0,5

1

1,5

2

2,5

3

Left lane TK Right lane TK

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 30 (33)


Figure 25. EN 13201 average surface luminance: January 23, 5pm, to January 24, 8am. Left lane (blue) and right lane (red), with traffic compensation.

Figure 26. Images taken at 9.00pm (top left) on 2016-01-23, 0.00am (top right), 3.00am (bottom left) and 6.00am (bottom right) on 2016-01-24. One out of two exposure times only.

0

0,5

1

1,5

2

2,5

3

Left lane TK Right lane TK

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 31 (33)


4.6. Extreme event analysis Spikes in the luminance signal may occur from time to time due to a number of reasons. One example is shown in Figure 27 where the uncompensated signal (shown in red colour) deviates strongly from the compensated one at approximately 9.40pm on the night between 2015-10-09 and 2015-10-10. The reason can be found in the images shown in Figure 28 taken just before, during and after the high-luminance event. The increase in luminance is caused by two or more vehicles illuminating the road surface. One is present within the field of view and one or two are just about to enter the measurement zone.

Figure 27. Luminance spike at around 9.40pm on the night between 2015-10-09 and 2015-10-10. Uncompensated Lu in red, compensated Lt in blue.

Figure 28. Images taken on 2015-10-09 at 9.23pm (left), 9.43pm (middle) and 10.03pm (right). One out of two exposure times only.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 32 (33)


5. Future work • A thorough analysis of the measurement error because of saturated pixels of the Lseq

camera and how it will influence the levels of Lth.

• Investigate the quality of vision and visibility of road markings under wet conditions. The road surface might appear brighter when wet, but the visibility of white lines may be poorer, so it is important that the road illumination system compensates for that.

• Investigate the influence of perspective error and specular reflexes that the Lseq camera may register when mounted significantly higher than the driver’s position.

• Analysis of the quality of vision in the tunnel opening during the day when the day light illumination changes. A comparison between tunnels with different design of the tunnel entrance portal would be interesting.

• A study of the sensitivity of the Lseq measure (as opposed to L20) to the camera rotation angle.

• A more thorough investigation into the angular dependence of road surface luminance estimation according to EN 132015.

• Explicit handling of specular surfaces such as pools of water through e.g. detection and masking.

• A field study looking at required sensor density (number of sensors per kilometre) to accurately model local conditions, with the objective of reaching a compromise allowing high-enough energy savings while maintaining traffic safety.

• A study of active lighting control in extreme conditions aiming to provide guidelines on how to control street lighting in e.g. foggy conditions and heavy rain/snow.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 33 (33)


6. References 1. MTk4P06467-2 Final report – Traffic compensated luminance estimation, 2015-06-24

2. ML150916-1rev1, Projektförslag Trafikkompenserad luminansmätning komplettering

3. R.V. Karandikar, J. Opt. Soc. Am. 45, (6), p483-8, (1955)

4. CIE 88:2004. ”Guide for The Lighting of Road Tunnels and Underpasses”, International Commission on Illumination, ISBN 3 901 906 31 2, 2nd edition

5. EN 13201:2003 Road lighting

SP Technical Research Institute of Sweden Measurement Technology - Communication Performed by

__Signature_1 __Signature_2

Mikael Lindgren

Appendices

Appendix 1: Road Surface Luminance Data – October 2015

Appendix 2: Road Surface Luminance Data – November 2015

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 1 (27)

Appendix 1


Appendix 1: Road Surface Luminance Data – October 2015 In this section, we report the average road surface luminance as a function of time at TKLM Test Site 2 where EN 13201 measurements have been performed. The measurements listed below are for October only. In each of the figures shown below, the unit on the y-axis is candela per square meter (cd/m2) and, on the x-axis, we show the time of day in hours from 6pm to 8am. Each figure shows measurements taken from one full night.

The sun set during October was between 4.28pm and 6.46pm. The sun rise, on the other hand, was between 7.16am to 7.23am.

In the figures below, the red line shows the uncompensated raw luminance measurements. The blue line, on the other hand, shows the traffic compensated luminance measurements.

In each of the figures, the summary statistics for the period is shown in the upper left corner. This includes the mean and median luminance values, the standard deviation of the luminance and the minimum/maximum luminance values during the time period. Note that the summary statistics have been provided for a shorter time period – 7pm to 7am.

In each of the figures, the upper plot is for the left lane and the lower plot for the right lane.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 2 (27)

Appendix 1


Figure A1- 1. EN 13201 average surface luminance: October 1, 6pm, to October 2, 8am.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 3 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 4 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 5 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 6 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 7 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 8 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 9 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 10 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 11 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 12 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 13 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 14 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 15 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 16 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 17 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 18 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 19 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 20 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 21 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 22 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 23 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 24 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 25 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 26 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 27 (27)

Appendix 1



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 1 (29)

Appendix 2


Appendix 2: Road Surface Luminance Data – November 2015 In this section, we report the average road surface luminance as a function of time at Test Site 2 where EN 13201 measurements have been performed. The measurements listed below are for November only. In each of the figures shown below, the unit on the y-axis is candela per square meter (cd/m2) and, on the x-axis, we show the time of day in hours from 4pm to 8 am. Each figure shows measurements taken from one full night.

The sun set during November was between 3.33pm and 4.25pm. The sun rise, on the other hand, was between 7.25am to 8.28am.

In the figures below, the red line shows the uncompensated raw luminance measurements. The blue line, on the other hand, shows the traffic compensated luminance measurements.

In each of the figures, the summary statistics for the period is shown in the upper left corner. This includes the mean and median luminance values, the standard deviation of the luminance and the minimum/maximum luminance values during the time period. Note that the summary statistics have been provided for a shorter time period – 5pm to 7am.

In each of the figures, the upper plot is for the left lane and the lower plot for the right lane.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 2 (29)

Appendix 2


Figure A2- 1. EN 13201 average surface luminance: November 1, 4pm, to November 2, 8am.

REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 3 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 4 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 5 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 6 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 7 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 8 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 9 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 10 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 11 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 12 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 13 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 14 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 15 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 16 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 17 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 18 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 19 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 20 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 21 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 22 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 23 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 24 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 25 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 26 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 27 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 28 (29)

Appendix 2



REPORT

Date Reference Page

2016-04-08 MTk5P08986-2 29 (29)

Appendix 2



Documents

Trafficcompensated luminance measurement complementary …fudinfo.trafikverket.se/fudinfoexternwebb/Publikationer... · 2016-12-22 · the intra-day variation of the visibility of